AI coding glossary
Prompt Engineering
Also known as: prompting, prompt design
In one sentence
The practice of writing inputs to LLMs to consistently get useful outputs, distinct from skills (instructions packaged as files) and fine-tuning (weight updates).
Full definition
Prompt engineering is the craft of structuring LLM inputs so the model produces consistently useful outputs. Core techniques: clear task framing, explicit output formatting, few-shot examples, role assignment, chain-of-thought prompting, and decomposition (breaking complex tasks into smaller prompted steps). In 2026 it overlaps heavily with SKILL.md authoring, a skill is a packaged prompt with structure (frontmatter for triggers, body for instructions). Industry view in 2026 is that prompt engineering matters less for frontier models than for older ones, Opus 4.7, GPT-5.5, and Gemini 3.1 Pro follow instructions reliably enough that elaborate prompting tricks (like 'think step by step') often add noise rather than signal. But structured prompts (clear sections, explicit output schemas, named roles) remain consistently useful.